This specification generally relates to systems and techniques for recognizing the words that a person is speaking, otherwise referred to as speech recognition.